Run a persistent OpenClaw Gateway on a DigitalOcean Droplet (~$6/month for the 1 GB Basic plan). DigitalOcean is the simplest paid VPS path. If you prefer cheaper or free options:Documentation Index
Fetch the complete documentation index at: https://docs.openclaw.ai/llms.txt
Use this file to discover all available pages before exploring further.
- Hetzner — €3.79/mo, more cores/RAM per dollar.
- Oracle Cloud — Always Free ARM (up to 4 OCPU, 24 GB RAM), but signup can be finicky and ARM-only.
Prerequisites
- DigitalOcean account (signup)
- SSH key pair (or willingness to use password auth)
- About 20 minutes
Setup
Create a Droplet
- Log into DigitalOcean.
- Click Create > Droplets.
- Choose:
- Region: Closest to you
- Image: Ubuntu 24.04 LTS
- Size: Basic, Regular, 1 vCPU / 1 GB RAM / 25 GB SSD
- Authentication: SSH key (recommended) or password
- Click Create Droplet and note the IP address.
Run onboarding
Access the Control UI
The gateway binds to loopback by default. Pick one of these options.Option A: SSH tunnel (simplest)Then open Then open Then open
http://localhost:18789.Option B: Tailscale Servehttps://<magicdns>/ from any device on your tailnet.Tailscale Serve authenticates Control UI and WebSocket traffic via tailnet identity headers, which assumes the gateway host itself is trusted. HTTP API endpoints follow the gateway’s normal auth mode (token/password) regardless. To require explicit shared-secret credentials over Serve, set gateway.auth.allowTailscale: false and use gateway.auth.mode: "token" or "password".Option C: Tailnet bind (no Serve)http://<tailscale-ip>:18789 (token required).Persistence and backups
OpenClaw state lives under:~/.openclaw/—openclaw.json, per-agentauth-profiles.json, channel/provider state, and session data.~/.openclaw/workspace/— the agent workspace (SOUL.md, memory, artifacts).
openclaw backup create is portable across hosts.
1 GB RAM tips
The $6 Droplet only has 1 GB RAM. To keep things smooth:- Make sure the swap step above is in
/etc/fstabso it survives reboots. - Prefer API-based models (Claude, GPT) over local ones — local LLM inference does not fit in 1 GB.
- Set
agents.defaults.model.primaryto a smaller model if you hit OOMs on large prompts. - Monitor with
free -handhtop.
Troubleshooting
Gateway will not start — Runopenclaw doctor --non-interactive and check logs with journalctl --user -u openclaw-gateway.service -n 50.
Port already in use — Run lsof -i :18789 to find the process, then stop it.
Out of memory — Verify swap is active with free -h. If still hitting OOM, use API-based models (Claude, GPT) rather than local models, or upgrade to a 2 GB Droplet.
Next steps
- Channels — connect Telegram, WhatsApp, Discord, and more
- Gateway configuration — all config options
- Updating — keep OpenClaw up to date